本论文关于K的公式如（2）式所示，它表示的是图结构中的邻域跳数。我们参

本论文关于K的公式如（2）式所示，它表示的是图结构中的邻域跳数。我们参看了在Indian Pines数据集上关于模型参数数量和模型处理的操作数量（Flops），其结果显示在下面的表1中。可以发现随着的增大模型的参数量保持不变，但是Flops逐渐变大。值得注意的是，实际上模型中（2）式的运算是稀疏运算，并不会产生这么大的计算量，但目前现有工具计算不了稀疏矩阵运算的Flops，故我们通过将（2）式转换为稠密矩阵运算得到的Flops以进行说明。

The formula for K in this paper is shown in equation (2), which represents the number of neighborhood hops in the graph structure. We have looked at the number of model parameters and the number of model processing operations (Flops) on the Indian Pines dataset, and the results are shown in Table 1 below. It can be observed that as the model increases, the number of parameters remains unchanged, but Flops gradually increases. It is worth noting that in reality, the operation of equation (2) in the model is sparse and does not generate such a large amount of computation. However, currently available tools cannot calculate the Flops of sparse matrix operation. Therefore, we will explain by converting equation (2) into Flops obtained from dense matrix operation.

The formula of K in this paper is shown in formula (2), which represents the number of neighborhood hops in the graph structure. We refer to the Indian Pines data set about the number of model parameters and the number of operations (Flops) of model processing, and the results are shown in Table 1 below. It can be found that the parameters of the model remain the same with the increase of, but the Flops gradually becomes larger. It is worth noting that, in fact, the operation of formula (2) in the model is a sparse operation, which will not produce such a large amount of calculation, but the existing tools can not calculate the Flops of sparse matrix operation, so we will explain it by converting formula (2) into the Flops obtained by dense matrix operation.